Implementing modified constraint-induced movement therapy after stroke in an early-supported discharge rehabilitation service: a process evaluation using RE-AIM QuEST

Abstract

Background

Modified constraint-induced movement therapy (mCIMT) improves upper limb (UL) function after stroke. Despite extensive evidence supporting its efficacy and multiple clinical practice guidelines recommending its provision, very few eligible stroke survivors receive mCIMT internationally. A multi-modal behaviour change intervention, guided by the Theoretical Domains Framework (TDF) and Behaviour Change Wheel, was implemented across a large early-supported discharge (ESD) rehabilitation service in Perth, Australia, to increase delivery of mCIMT within existing resources. As part of this iterative implementation process, qualitative data from stroke survivors, carers, clinicians and managers informed program adaptations to support sustained mCIMT delivery. This study aimed to evaluate the process of implementing mCIMT across the seven sites of this ESD service.

Methods

This mixed methods process evaluation was guided by the RE-AIM QuEST framework. Data collection included patient file audits (conducted twice over six-month periods) (N = 540), clinician training attendance records, interviews with stroke survivors and carers (N = 25), and focus groups with clinicians and managers (N = 24). Quantitative data were analysed descriptively. Qualitative data were analysed using reflexive thematic analysis and mapped to the Capability, Opportunity, Motivation- Behaviour model for interviews and TDF for focus groups.

Results

Reach: 31% of all stroke survivors admitted to the ESD service, including 80% of those with an UL impairment, were eligible for mCIMT. Twenty-eight facilitated workshops were conducted across all sites (four per site). Effectiveness: Clinicians reported that the behaviour change intervention improved their knowledge, skills and confidence to identify suitable stroke survivors and deliver mCIMT programs. Adoption: 17% of eligible stroke survivors were offered mCIMT and 11% received it (Audit 1). Implementation: 7/9 programs were delivered with protocol fidelity. Adaptations included ongoing/booster training sessions and resource refinement. Maintenance: 34% of eligible stroke survivors were offered mCIMT, with 26% receiving it, and all programs adhered to the protocol (Audit 2).

Conclusions

mCIMT was feasibly and acceptably integrated into the ESD service, demonstrating that complex interventions can be implemented within existing resources. However, concerns persist regarding sustained delivery. Future research should investigate long-term sustainability and target policy-level change to ensure stroke survivors consistently receive evidence-based interventions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12913-025-13273-9.

Contributions to the literature

• Modified constraint-induced movement therapy (mCIMT) is a highly effective intervention to improve upper limb function after stroke, yet very few stroke survivors receive it.

• The use of co-designed, multi-modal, theory-informed, behaviour change interventions, enabled the successful integration of mCIMT in an early-supported discharge rehabilitation service without requiring additional resources.

• Our process evaluation highlights the feasibility of implementing evidence-based practices in resource-constrained health systems; however, sustainability remains a challenge.

• Future research should focus on strategies to sustain mCIMT delivery amid increasing health system demand so that all stroke survivors can receive best practice rehabilitation.

Introduction

Upper limb (UL) impairment after stroke reduces independence in activities of daily living, increasing burden of care and reducing quality of life [1]. Constraint-induced movement therapy (CIMT), including modified versions (mCIMT), is the most researched and best supported rehabilitation intervention to improve UL function after stroke [2–4]. Findings from multiple systematic reviews support its efficacy [4–7], and several national clinical practice guidelines for stroke management recommend that clinicians should provide CIMT/mCIMT to eligible stroke survivors [8–10]. Despite these strong recommendations, there remains a significant evidence-to-practice gap, with few stroke survivors receiving this intervention [2, 11, 12]. This gap exemplifies a broader issue in stroke rehabilitation, where evidence-based practices are often not implemented into routine clinical care [13, 14].

CIMT/mCIMT is a complex intervention with multiple interdependent components operating concurrently as part of an overall treatment package [15, 16]. Three main components make up the CIMT/mCIMT package: intensive, task-oriented practice using the more-affected UL, consisting of repetitive task practice and shaping; a suite of behavioural strategies that aim to promote use of the more-affected UL outside of therapy, known as the transfer package; and restraint of the less-affected UL for periods of the day [15, 17]. The transfer package is the most important component to CIMT programs [18] however it is often omitted from CIMT/mCIMT programs in clinical settings [11, 19, 20]. Implementation of complex interventions like CIMT/mCIMT is influenced by a range of determinants within healthcare systems, including organisational, clinician, and patient factors [21]. Prior research has identified key barriers to the routine delivery of CIMT programs, such as clinicians’ lack of knowledge, skills and confidence to deliver these programs, as well as perceived constraints relating to staffing, time, and resources [22, 23]. Clinicians have also reported their own concerns about patients’ ability to tolerate the intensity of mCIMT, further limiting its implementation [23]. These barriers are consistent across diverse clinical settings and countries [23], suggesting that attempts to address them could have wide-ranging impact across the spectrum of stroke rehabilitation.

In response to minimal provision of mCIMT identified through routine auditing, a co-designed implementation strategy was initiated within Rehabilitation in the Home (RITH), a large, publicly-funded (Government), early-supported discharge (ESD) rehabilitation service in Perth, Australia. RITH comprises seven sites across the entire Perth metropolitan region, providing allied health rehabilitation in a generalist capacity for all diagnostic groups. Approximately 700 stroke survivors are referred to RITH annually for home-based rehabilitation following discharge from hospital. An earlier attempt in 2017 to increase mCIMT delivery through clinician education and training was unsuccessful due to a range of previously reported factors [24]. To address these limitations, a multi-modal, theory-informed, behaviour change intervention was developed and delivered (commenced in July 2018) [24]. This co-designed approach engaged clinicians (physiotherapists, occupational therapists and therapy assistants) to understand their current practice, clarify their perceived importance of providing evidence-based UL rehabilitation, understand their knowledge around delivering mCIMT programs, and to identify how mCIMT could be adapted to fit within the current service delivery model without additional resources, staffing or funding [25]. Determinants of implementation for mCIMT, which mirrored those reported in prior literature, were then mapped to the Theoretical Domains Framework (TDF) [26]. Strategies to address these determinants were guided by the Behaviour Change Wheel (BCW) [27] to develop the behaviour change intervention. An implementation package was then delivered which consisted of extensive education and staff training, expert peer modelling, development of context-specific resources, and championing interdisciplinary collaboration [24]. This iterative process then used qualitative data obtained from stroke survivors and carers [28], and clinicians and managers [29], to inform adaptations to the mCIMT program to support sustained delivery of mCIMT (Fig. 1). The implementation program was delivered alongside routine clinical duties, with time for development, delivery, and adaptation (P3 clinicians), as well as participation in training and focus groups (clinicians and managers), accommodated through individual workload management rather than dedicated infrastructure or additional resourcing.

Fig. 1.

Timeline of implementation process

Given the scale and complexity of this implementation program, a process evaluation was warranted to examine how the intervention was delivered, identify further contextual influences, and understand the mechanisms underpinning its success or failure. The use of comprehensive, theory-informed frameworks provide structure to determine if the active components of an implementation intervention led to uptake of the evidence-based practice. The RE-AIM framework assesses five key domains [30]: Reach (how well the intervention reaches the target population), Effectiveness (stakeholders’ perceptions of the intervention’s effectiveness), Adoption (uptake of the intervention by clinicians), Implementation (program adaptations and fidelity to the protocol), and Maintenance (sustainability of intervention delivery over time). The RE-AIM Qualitative Evaluation for Systematic Translation (QuEST) framework expands on the original RE-AIM framework by incorporating qualitative questions within each of the RE-AIM dimensions, providing deeper insights into the underlying reasons behind the intervention’s success or failure [31]. Using this mixed methods approach enables a more in-depth retrospective evaluation to understand the reasons for an intervention’s success or failure, while also allowing researchers to adapt and refine the intervention during its implementation to ensure greater fidelity and consistency across sites [31].

The overarching aims of this implementation program were to determine if mCIMT was feasible, acceptable, and sustainable in the RITH service without additional staffing, resources or funding. Specifically, this study aimed to evaluate the process of implementing mCIMT across all seven sites of the RITH service, with research questions mapped to the RE-AIM QuEST framework.

1. Reach.

   1. What proportion of stroke survivors were eligible to receive mCIMT in the RITH service?
   2. Did the behaviour change intervention reach all relevant clinicians? Were there barriers impacting the reach of the behaviour change intervention to all clinicians?

2. Effectiveness.

   1. How did clinicians perceive the effectiveness of the behaviour change intervention?

3. Adoption.

   1. What proportion of eligible stroke survivors were offered and received mCIMT as part of their UL rehabilitation? Was there variability in adoption between sites?
   2. What barriers and enablers influenced adoption of mCIMT?

4. Implementation.

   1. Did the mCIMT programs delivered by RITH clinicians adhere to correct protocols?
   2. What adaptations were made to the mCIMT program to increase delivery of the intervention?

5. Maintenance.

   1. Was the proportion of mCIMT programs delivered to stroke survivors maintained or improved at a follow-up data collection point (July to December 2021)?
   2. Did the mCIMT programs delivered adhere to correct protocols at the follow-up data collection point?

Methods

This single cohort descriptive study used mixed methods to evaluate the process of implementing mCIMT across the RITH service. The Standards for Reporting Implementation Studies (StaRI) [32] and the Template for Intervention Description and Replication (TIDieR) [33] checklists were followed to ensure comprehensive reporting (Supplementary Files 1 and 2). This study was registered on the Australian New Zealand Clinical Trials Registry as an observational study (ACTRN12620000079943, registered 30th January 2020).

Setting

The implementation project and process evaluation involved all seven sites of the RITH service. RITH is a publicly-funded, ESD rehabilitation service operating across the Perth metropolitan region. RITH provides coordinated allied health rehabilitation and accepts referrals from all publicly-funded hospitals and inpatient wards. During the mCIMT implementation period, RITH experienced substantial growth due to the COVID-19 pandemic, as demand for home-based healthcare services increased to free up hospital capacity for patients with acute respiratory conditions throughout 2020. During 2021, RITH launched a Discharge-to-Assess model in collaboration with several public hospitals, which facilitated early discharge of older patients (> 65 years of age) for allied health assessments at home. Referral numbers grew substantially during this period without corresponding staffing increases. Table 1 provides information on referral numbers, occasions of service, and staffing levels over the study period.

Table 1.

RITH service data for referrals, occasions of service and staffing over the implementation period

Financial year	Total number of referrals	Total occasions of service	Staffing FTE				Annual growth- referrals	Annual growth- staffing
			Total staff	PT	OT	TA
2018-19	5,954	45,795	78.19	20.3	10.1	7.6
2019-20	6,464	48,226	79.66	20.6	11.2	8.4	9%	1.9%
2020-21	7,614	50,854	80.76	23	12.6	9	18%	1.4%
2021-22*	8,889	60,062	99.47	27.6	14.4	12.2	17%	23.1%

FTE- full time equivalent; PT- physiotherapists; OT- occupational therapists; TA- therapy assistants

Note: Financial years run from July 1st to June 30th

^*In late 2021, funding approval for an additional 21.5 FTE staffing occurred, and recruitment to these positions began in early 2022, which was after the completion of Audit 2 data period

Participants

This study involved clinicians, allied health managers, stroke survivors and carers throughout all stages of the implementation program. Detailed descriptions of each stakeholder group are published elsewhere [24, 28, 29].

P3 neurological physiotherapist and occupational therapist

The P3 Neurological Physiotherapist and Occupational Therapist(s) led the implementation and provided peer support, training, and resource development. The P3 classification in Western Australia Health is an advanced level of practice, signifying extensive clinical experience in the respective area of practice, skills in clinical supervision, and possession of a relevant post-graduate qualification. Their roles have been previously described in detail [24, 29].

At the commencement of this project, the RITH service had one full-time P3 Neurological Physiotherapist (1.0 full time equivalent (FTE)), and two part-time P3 Occupational Therapists (1.0 FTE total). From May 2019, one Occupational Therapist left the service, and the position was not filled, resulting in one occupational therapist working 3 days per week (0.6 FTE) for the remainder of the project. Initially, the P3 clinicians provided support across five RITH sites (Hubs B and C), but from February 2019, their roles expanded to include all seven sites, rotating between hubs every four months (Table 2). These rotations allowed the P3 clinicians to offer support across all sites, however geographic constraints and increased caseloads limited their ability to provide support for clinicians not co-located at the same site as the P3.

Table 2.

P3 physiotherapist and occupational therapist rotation schedule (annual)

Period	Hub A	Hub B	Hub C
February–May	P3 Physiotherapist	–	P3 Occupational therapist
June–September	P3 Occupational therapist	P3 Physiotherapist	–
October–January	–	P3 Occupational therapist	P3 Physiotherapist

Hub A- based at Sites 1 and 2; Hub B- based at Site 3, additional support to Sites 4 and 5; Hub C- based at Site 6, additional support to Site 7

The P3 clinicians’ involvement in this project was in addition to their usual roles, which included clinical consultation and supervision, development and delivery of continuing education, and liaison/in-reach with referring hospitals.

Clinicians and managers

All physiotherapists, occupational therapists and therapy assistants of the RITH service were included in this implementation project. For the qualitative study, purposive sampling was undertaken to recruit a representative sample of clinicians across the service, and convenience sampling was undertaken to recruit managers, which has been described in detail previously [29].

Stroke survivors and carers

Patients who received RITH physiotherapy and/or occupational therapy after stroke were included for retrospective auditing to determine eligibility and adoption of mCIMT. Targets for successful implementation were set at 75% of eligible stroke survivors being offered mCIMT, with at least 50% receiving the intervention. These targets were based on data published by Christie and colleagues following their CIMT implementation study, conducted in Sydney, Australia between 2016 and 2019 [34].

Inclusion criteria for cohort audit:

• Primary diagnosis of unilateral stroke (ischaemic and/or haemorrhagic) as defined by hospital discharge summary, at any time-point (acute, subacute, chronic).

• Referred for physiotherapy and/or occupational therapy.

Eligibility criteria for mCIMT were determined by:

• Documented UL impairment from stroke.

• Documented ability to extend wrist and fingers (score of > 1 on the hand domain of the Upper Limb sub-scale of the Motor Assessment Scale).

• No significant cognitive impairment (i.e.: Montreal Cognitive Assessment > 24 or similar) or severe communication impairment impacting ability to provide informed consent to participate in mCIMT.

• No evidence of untreated mood disturbance or depressive disorder (i.e.: Patient Health Questionnaire-2 < 3).

• No/minimal documented shoulder pain on stroke-affected side.

• Documented presence of carer or family member to support mCIMT program (Note: not having a carer at home did not exclude a patient from participating in mCIMT; however, it may have influenced the treating therapist’s decision to offer the intervention).

Purposive sampling was used to recruit a diverse range of stroke survivors and carers to participate in semi-structured interviews for the qualitative study. These individuals either underwent or supported a mCIMT program during their admission with RITH. Sampling sought diversity in respect to participant age, sex, type of stroke, hand dominance, time post-stroke (early subacute vs. late subacute), RITH base, and relationship to stroke survivor among carers. Recruitment details have been described previously [28].

Data collection and analysis

Data were collected across the RE-AIM QuEST domains from various sources, including:

• Medical file audits.

• Training attendance records.

• Semi-structured interviews with stroke survivors and carers.

• Focus groups with therapists, therapy assistants and managers.

Quantitative data

Training attendance records were kept for all mCIMT specific training sessions conducted by the P3 clinicians. Retrospective audits were conducted over two six-month time periods, separated by a six-month period during which adaptations to the mCIMT program based on qualitative feedback from stakeholders were applied. The first audit, originally scheduled for early 2020, was postponed by six months due to the COVID-19 pandemic. Consequently, the audits were conducted using patient files for stroke survivors admitted to the RITH service from July 2020 to December 2020 (Audit 1) and July 2021 to December 2021 (Audit 2). These audits captured:

• Patient demographic data (age, gender, stroke, date of stroke, hospital length of stay, hand-dominance, pre-stroke employment status, and presence of partner/carer).

• Eligibility for mCIMT (UL impairment, absence of cognitive or severe communication deficit, adequate cognitive screen scores (if performed), and minimal/absent shoulder pain).

• Whether eligible stroke survivors were offered and received mCIMT, and adherence to the mCIMT protocol (to evaluate intervention fidelity).

• Adherence to the protocol was determined by the documented presence of:

  • A mCIMT program which was a minimum two-weeks duration with at least three sessions per week (minimum six training sessions).
  • All elements of the Transfer Package including behavioural contract, regular administration of the Motor Activity Log (MAL), home skills assignment tasks recorded in a training diary, and problem-solving with clinicians.
  • Intensive task-oriented practice including shaping and repetitive task practice.
  • Restraint use or documented discussion on how to increase use of more-affected UL if not using a restraint.

Descriptive statistics were used to analyse quantitative data, and proportions were calculated to assess mCIMT adoption.

Qualitative data

Following the delivery of the behaviour change intervention and during the first audit period, qualitative data were collected through semi-structured interviews with stroke survivors and carers and focus groups with clinicians and managers. Interviews and focus groups were audio-recorded, transcribed verbatim and uploaded into NVivo12 [35] for analysis.

Interviews were analysed using reflexive thematic analysis [36] and deductively mapped to the Capability, Opportunity, Motivation- Behaviour (COM-B) model [28]. Data from focus groups underwent reflexive thematic analysis, with themes deductively mapped to the TDF [29]. Both studies aimed to explore the acceptability of mCIMT in the RITH setting and to identify adaptations required to support sustained delivery, with detailed methods and the justification for selection of these frameworks published previously [28, 29].

Results

Reach

What proportion of stroke survivors were eligible to receive mCIMT in the RITH service?

The proportion of stroke survivors who met criteria for mCIMT is presented in Table 3. At the time of this study, six sites used paper-based medical files, housed in different facilities with varying accessibility for audit. While attempts were made to audit files for all stroke survivors admitted during the audit periods, some files (n = 167) were unavailable for review.

Table 3.

Proportion of stroke survivors eligible for mCIMT

Audit period	Stroke survivors admitted to  RITH for PT and/or OT N	Files audited N (%)	UL assessed N (%)	UL impairment  identified N (%)*	Eligible for  mCIMT N (%)#
July–Dec 2020	325	237 (73)	183 (77)	103 (56)	82 (35)
July–Dec 2021	382	303 (79)	290 (96)	107 (37)	86 (28)
Total	707	540 (76)	473 (88)	210 (44)	168 (31)

mCIMT- modified constraint-induced movement therapy; PT- physiotherapy; OT- occupational therapy; RITH- Rehabilitation in the Home; UL- upper limb

# Calculated from total number of files audited

* Calculated from number of UL assessed

31% of all stroke survivors in this cohort were eligible for mCIMT. RITH accepts referrals for stroke survivors with a range of impairments and functional limitations, and not all stroke survivors referred have UL impairment. 80% of those with an UL impairment were eligible for mCIMT.

Did the behaviour change intervention reach all relevant clinicians? Were there barriers impacting the reach of the behaviour change intervention to all clinicians?

Twenty-eight facilitated behaviour change workshops were conducted throughout 2019: four at each of the seven RITH sites. Small group workshops allowed for greater experiential and practical learning for clinicians. Delivery of multiple face-to-face workshops increased the number of training sessions available to train part-time staff, who could attend a session at another site if required. Across the four facilitated workshops, 46 clinicians attended session one, 40 attended session two, 45 attended session three, and 43 attended session four. The aims of each workshop and learning content have been previously described [24].

Focus group data collected from therapists, therapy assistants and managers identified several barriers to reaching all relevant clinicians and have been described in detail previously [29]. The key determinants of implementation (and associated TDF domains) are outlined below:

Barriers:

• Difficulty for part-time staff to attend training sessions (skills).

• Inequity in access to the P3 clinicians for peer modelling (social influences).

• Perceived lack of suitable stroke survivors to develop mCIMT-specific skills (skills).

• Time and staffing constraints to attend training and observe P3 clinician modelling (environmental context and resources).

• Difficulties with face-to-face delivery of mCIMT during the first six months of 2020 due to the COVID-19 pandemic (environmental context and resources).

The strategies adopted to address barriers to reaching all clinicians are reported under Implementation.

Effectiveness

How did clinicians perceive the effectiveness of the behaviour change intervention?

Qualitative data obtained from clinicians identified that the behaviour change intervention had improved their knowledge, skills and confidence to identify suitable stroke survivors and deliver mCIMT programs within the RITH service. These perceptions have been described in detail previously [29].

Adoption

What proportion of eligible stroke survivors were offered and received mCIMT as part of their UL rehabilitation? Was there variability between sites for adoption of mCIMT?

The proportion of eligible stroke survivors who were offered and subsequently received mCIMT are presented in Table 4.

Table 4.

Summary of mCIMT eligible, offered and received by each RITH base- audit 1

Hub	RITH site	Total stroke survivors	Files audited	Eligible	Offered N (%)	Received N (%)
Hub A	Site 1	56	33	7	1 (14)	0 (0)
	Site 2	32	24	6	2 (33)	1 (17)
Hub B	Site 3	60	46	18	1 (6)	1 (6)
	Site 4	50	22	10	1 (10)	0 (0)
	Site 5	40	30	9	0 (0)	0 (0)
Hub C	Site 6	48	43	18	7 (39)	5 (28)
	Site 7	39	39	14	2 (14)	2 (14)
	Total	325	237	82	14 (17)	9 (11)

What barriers and enablers influenced staff adoption of mCIMT?

Several new determinants influencing the adoption of mCIMT were identified during the focus group study conducted with clinicians and managers. These have been described in detail elsewhere [29], however a brief outline of these determinants (with associated TDF domains) are listed below.

Barriers:

• Difficulty balancing other rehabilitation goals with mCIMT provision (goals).

• Time and staffing constraints (environmental context and resources).

• Inequity in access to P3 neurological therapist for clinical support (social influences).

• Therapist cognitive overload and fatigue from changes to multiple systems and service processes (memory, attention, and decision processes).

• Lack of patient awareness of mCIMT (knowledge).

• Inconsistent therapy resources between sites to run mCIMT sessions (environmental context and resources).

Enablers:

• Clear understanding about the individual components that make up the mCIMT package (knowledge).

• Interdisciplinary collaboration/shared workloads (beliefs about capabilities).

• The rich opportunities provided by the home environment to deliver mCIMT (knowledge).

• Standardisation of practice area which increased consistency between clinicians and saved time (belief about capabilities).

• Thinking ‘mCIMT first’ (attention, memory, and decision processes).

Implementation

Did the mCIMT programs delivered by RITH clinicians adhere to correct protocols (intervention fidelity)?

In Audit 1, adherence to the RITH mCIMT protocol was inconsistent. While all nine programs included shaping, only seven programs fully adhered to the protocol. Five of these programs involved a physiotherapist (PT), occupational therapist (OT) and therapy assistant (TA), while two involved an OT and TA. The two programs that did not adhere to the protocol were delivered solely by an OT, and lacked key components such as the behavioural contract, intensive task practice and consistent administration of the MAL. Neither of these two programs met the duration (at least two weeks) or frequency (three times per week) requirements.

What adaptations were made to the mCIMT program to increase delivery of the intervention?

Several adaptations were made to the mCIMT program based on feedback gained from stakeholders. Table 5 outlines the strategies suggested by the stakeholders with comments on which adaptations were implemented.

Table 5.

Adaptations made to the mCIMT program in response to stakeholder feedback

Suggested implementation adaptation/strategy	Change made/comment
Debrief session requested by stroke survivors upon completion of mCIMT program	Additional therapy session scheduled at end of mCIMT program to allow stroke survivors to reflect on their progress, review outcome measures, celebrate successes, and make plans to ensure momentum is maintained once structured therapy has ceased
Regular booster and on-demand training	Recorded online content for theory component of mCIMT program uploaded onto Education channel of RITH Microsoft Teams site for on-demand access Regular, six-monthly practical sessions conducted at two RITH sites to capture new and/or inexperienced staff. Additional focus on the importance of observing P3 clinician run a program in its entirety (modelling) to complete training process
Consistent mCIMT therapy box at each site	Purchase of two mCIMT boxes per site, containing a wide range of therapy and everyday resources which could be used to perform shaping tasks. All sites had consistent resources and a designated OT responsible for keeping all items intact, or order when items had been used/lost
mCIMT booklet	Modifications to paper-based resources and compiled into single, printable PDF file, with support from South Metropolitan Health Service Publications team and the consumer advisory group
Add mCIMT to learning objectives for PT and OT	New staff are expected to complete a range of learning objectives within the first three months of employment in RITH. mCIMT was added to the learning objectives as advanced skills, requiring the new staff member to contact the P3 clinician and enquire when the next education and training sessions for mCIMT would run
Site champions	Site champions flagged for each RITH site; however, this did not proceed due to the RITH service requiring multiple champions per site for implementation of Microsoft Office 365 applications and roll out of the RITH App, which would manage referrals, data collection and workflows
Audit and feedback	Audit and feedback were conducted on one occasion following Audit 1. Regular audit and feedback sessions were intended, however logistical and time constraints negated the delivery of regular audit and feedback. This strategy was not undertaken as planned
Flag for mCIMT on intake	Admissions therapists (intake officer) were advised to query all stroke referrals for eligibility for mCIMT and flag on patient file for treating therapist
Education sessions at hospital to support building stroke survivor engagement for mCIMT upon admission to RITH	Professional development sessions conducted at several inpatient rehabilitation and community outpatient services educating on the RITH mCIMT program, eligibility criteria for mCIMT, and requests for referring clinician to flag for mCIMT when making the referral to RITH
Prompts on stroke patients’ identifier to cue discussion regarding eligibility for mCIMT at team meetings	Clinicians were advised to add a ‘mCIMT’ written prompt on a patient’s whiteboard strip for discussion during each RITH site’s multidisciplinary team meeting. This strategy was adopted at two sites, but not at the other five

mCIMT- modified constraint-induced movement therapy; PT- physiotherapist; OT- occupational therapist; RITH- Rehabilitation in the Home

Maintenance

Was the proportion of mCIMT programs delivered to stroke survivors maintained or improved at a second data collection point?

The proportion of eligible stroke survivors who were offered and received mCIMT increased during the period of Audit 2 compared with the first audit data (Table 6).

Table 6.

Summary of mCIMT eligible, offered and received by each RITH base- audit 2

Hub	RITH site	Total stroke survivors	Files audited	Eligible	Offered (%)	Received (%)
Hub A	Site 1	71	41	11	1 (9)	0 (0)
	Site 2	28	18	4	2 (50)	1 (25)
Hub B	Site 3	60	50	16	6 (38)	5 (31)
	Site 4	50	44	7	2 (29)	2 (29)
	Site 5	52	45	13	7 (54)	5 (39)
Hub C	Site 6	66	61	19	5 (26)	5 (26)
	Site 7	55	44	16	6 (38)	4 (25)
	Total	382	303	86	29 (34)	22 (26)

Were subsequent programs delivered in adherence to the protocol?

Adherence to the protocol improved in Audit 2, with all 22 delivered programs meeting the protocol requirements. All programs included a behavioural contract, regular use of the MAL, home skills assignment, shaping, intensive task practice, and restraint use. Every program met the duration and frequency requirements set out in the RITH mCIMT protocol. Eighteen programs involved a PT, OT and TA, three programs involved an OT and TA, and one program involved a PT and OT.

Discussion

This study evaluated the process of implementing mCIMT across a large, multi-site, ESD rehabilitation service. By using multi-modal, theory-informed behaviour change strategies, guided by co-design and stakeholder engagement, mCIMT was feasibly and acceptably integrated into clinical practice without access to additional funding, resources, or staffing. This demonstrates the feasibility of delivering mCIMT in resource-constrained health services, particularly in ESD settings where limited prior research exists. Our findings suggest that mCIMT can be effectively delivered in this context, highlighting its potential as a standard intervention for UL rehabilitation among eligible stroke survivors. However, challenges persist regarding the sustained delivery and broader adoption of the intervention.

Following the initial behaviour change intervention, a small increase in mCIMT delivery was observed in Audit 1. However, despite the extensive training, support and resource development, only 11% of eligible stroke survivors received mCIMT, which was well below the implementation target of 50%. Several factors contributed to this low uptake. First, the implementation period was interrupted by the COVID-19 pandemic, which caused delays in data collection and disruptions to routine service delivery. The loss of momentum and enthusiasm for delivery of mCIMT, combined with shifting priorities to service provision during and after community lockdowns, likely influenced adoption rates. Additionally, staffing turnover during this period meant that many new clinicians did not receive training or exposure to mCIMT. In response, several strategies were implemented including delivery of regular booster training sessions, further refinement of resources, and formalising mCIMT into the learning competencies that new staff undertake during onboarding to the service. Improvements in mCIMT delivery were observed at most sites (Audit 2) to approximately one-third of eligible stroke survivors being offered mCIMT, and 26% receiving it. However, with only two audit points, it remains unclear whether these strategies could continue to increase delivery towards the 50% implementation target.

Access to P3 neurological clinicians played a critical role in supporting mCIMT delivery during this implementation project. P3 clinicians were valued for their ability to support identification of suitable patients, model the intervention, and offer clinical problem-solving support [24, 29]. However, inequitable access to P3 support was identified as a barrier to mCIMT delivery, particularly for hubs and sites where the P3 was not co-located. Due to annual rotations, RITH hubs had several months without P3 support each year, as seen in Hub A during both audit periods (Table 2). This lack of consistent access to clinical expertise, combined with logistical challenges such as limited car availability for joint home visits, contributed to the variability in mCIMT delivery across sites. Site-based facilitation enhances implementation outcomes in rehabilitation settings [34, 37–39], and the absence of this support likely impacted Hub A’s uptake of mCIMT, resulting in far fewer mCIMT programs being delivered at these sites compared to those at other hubs (Tables 4 and 6).

The reliance on the P3 clinicians as the primary champions of mCIMT also emerged as a potential threat to long-term sustainability [29]. Site champions were proposed as a solution to mitigate this [24, 29], but competing service priorities hindered their recruitment. During 2021, the RITH service began work on development and implementation of a service-wide information and communication technology (ICT) solution for managing referrals, workflows, data collection and patient flow through the service. This large-scale project required multiple clinical champions per site to upskill in ICT systems and train clinicians using a ‘train-the-trainer’ model. Many of the clinicians flagged as mCIMT champions were instead recruited as champions for this ICT project. The absence of local champions presents a continuing challenge for the sustained and equitable delivery of mCIMT across the service. This highlights a fundamental lesson: implementation of complex interventions like mCIMT require careful negotiation with service leaders to align the demands of implementation with existing and future service priorities. Future implementation efforts must account for these competing demands on service delivery by ensuring proposed interventions can be adopted without overwhelming clinical resources.

Intervention fidelity is important for clinicians delivering evidence-based interventions such as mCIMT, as deviations from the specified protocol can reduce its effectiveness in real-world practice [40]. For example, forced use therapy, which involves applying a restrictive mitt to the less-affected UL without including shaping and the transfer package, has shown to confer little to no benefit in comparison to control interventions [7]. Similarly, a randomised controlled trial by Baldwin and colleagues [19] found no significant difference in outcomes between mCIMT and usual care occupational therapy, likely due to the mCIMT group receiving only the mitt-wearing component without the inclusion of shaping or transfer package components [18]). These findings underscore the importance of delivering all components of the mCIMT protocol [17, 18]. A key aspect of the behaviour change intervention was to ensure that RITH mCIMT programs were delivered with strict adherence to the protocol [24], emphasising the importance of the transfer package to facilitate long term behaviour change in stroke survivors [18]. Most of the mCIMT programs delivered during Audit 1 and all programs delivered during Audit 2 incorporated all components of the mCIMT package. The two programs that did not adhere in Audit 1 were delivered by individual clinicians, highlighting the importance of interdisciplinary collaboration to ensure the feasibility of delivering mCIMT programs with fidelity. The present findings demonstrate that delivery of mCIMT is feasible within existing resources and staffing, if delivered collaboratively by physiotherapists, occupational therapists and therapy assistants.

A common perception among clinicians is that few stroke survivors are suitable candidates for mCIMT, with eligibility often suggested to be around 10% [7]. The origin of this figure is unclear, though prior research suggests an eligibility range between 3 and 22% [41], which is based on screening for clinical trial participation and may not be reflective of real-world data for stroke survivors accessing rehabilitation services. Our data challenge this perception, as we identified 31% of stroke survivors referred to RITH were eligible for mCIMT. In addition to volitional UL motor activity, we also considered confounders such as cognitive impairment, mood, availability of carer support, and presence of severe language impairment, which were reflective of the clinical reasoning processes undertaken by therapists in determining suitability for mCIMT. Our findings were consistent with those published by Christie and colleagues, who reported 32% eligibility in their large CIMT implementation study conducted in Sydney, Australia [34]. Further, specifically for RITH stroke survivors with an UL impairment, 80% were eligible for mCIMT. ESD services tend to provide rehabilitation to a greater proportion of those with mild to moderate stroke presentations [42], which may explain the high proportion of eligible stroke survivors in our data. This high proportion has not been reported previously and provides a strong argument for mCIMT to become the standard UL intervention in the ESD setting. Our data, combined with data from Christie and colleagues’ [34], provide real-world evidence that a substantial proportion of stroke survivors could benefit from this highly effective intervention. These findings must be disseminated widely, including within clinical practice guideline recommendations, to challenge existing perceptions that (m)CIMT is only suitable for a limited subset of stroke survivors.

Time constraints are frequently cited as a barrier to implementing complex rehabilitation interventions [23, 25, 43, 44]. The RITH mCIMT protocol was developed with input from clinicians, aiming to fit within the time and staffing constraints of the service as it was in 2018. However, over the course of this study, referral numbers increased by 49%, with only a 3% increase in staffing by the end of the data collection period (noting that the major increase in staffing for the 2021/22 financial year occurred after completion of Audit 2). In addition, several new initiatives had been launched across RITH targeting service growth and ICT transformation, along with the lingering influence of COVID-19, which placed further cognitive and workload demands on clinicians. While the impact of these pressures on mCIMT delivery is unclear, they likely contributed to our lower-than-expected adoption rates and must be considered when evaluating the overall success of the implementation program.

The lack of audit data points limited our ability to assess changes in adoption rates over time and to thoroughly evaluate long-term maintenance. We were unable to undertake regular and frequent audit and feedback, which is an effective method of supporting behaviour change in clinical settings [34, 38, 39, 45, 46]. While feedback was provided following Audit 1, this occurred only once, rather than on a frequent basis as originally intended. Challenges in accessing patient medical files, which were distributed across eight locations with varying levels of access, further compounded the difficulty of conducting timely audits. Additionally, audit data collection was solely undertaken by the P3 Neurological Physiotherapist (AW), whose workload increased significantly in 2019 as the number of supported sites expanded from five to seven. Concurrently, the P3 Neurological Occupational Therapist staffing decreased from 1.0 FTE to 0.6FTE, requiring the P3 Neurological Physiotherapist to also support more occupational therapists with UL rehabilitation programs. The introduction of new RITH programs, such as Discharge-to-Assess, necessitated recruitment of staff with acute medical care experience, rather than a background in neurological rehabilitation. These staff required greater P3 support for stroke rehabilitation, increasing the frequency of joint consultations provided by P3 clinicians with stroke patients and adding to their clinical workload. These constraints limited our ability to perform more frequent audits and may have reduced the overall effectiveness of the implementation program. Nonetheless, these challenges demonstrate the pragmatic realities of implementing mCIMT in a large, multi-site ESD service, particularly considering the broad impact of the COVID-19 pandemic concurrent with this project.

Furthermore, our evaluation of the implementation program did not allow for direct comparison of particular behaviour change strategies, and the absence of a control group limits our ability to draw definitive conclusions about the specific factors driving mCIMT uptake. While trialling different strategies at different sites could have helped identify the most effective interventions, this was beyond the scope of this pragmatic, clinician-led project. Evidence suggests that successful implementation strategies are multifactorial [39], meaning that isolating and testing individual strategies may not have been appropriate in this context. Without a control group, the complexity of the implementation process and the interplay of various strategies make it difficult to attribute success to any single intervention.

Despite the reported challenges, this study has several strengths. First, we successfully implemented mCIMT into routine clinical practice without additional resources, at a time when the RITH service experienced significant growth and was engaged in multiple other service-wide initiatives, including navigating the complexities of the COVID-19 pandemic. This highlights the potential for integrating complex, evidence-based rehabilitation interventions in real-world settings despite competing priorities. Second, the use of multiple implementation science frameworks, informed by stakeholder engagement and co-design, was central to designing and evaluating effective intervention strategies for implementing mCIMT in the RITH service and is another strength of this study. The TDF guided the identification of key behavioural determinants from clinician and manager perspectives, such as capability gaps, time constraints, and social influences, which evolved as clinicians gained experience with mCIMT and faced new challenges to rehabilitation service delivery. Using the COM-B model to analyse qualitative data from stroke survivors and carers provided insights into how their capability, opportunity, and motivation influenced engagement with the mCIMT program. This analysis identified that carers wanted to support the program, and that both stroke survivors and carers needed better understanding of the intervention to increase their engagement. The BCW enabled the selection of appropriate intervention functions, such as on-demand education, regular training, and environmental restructuring, to address barriers and champion enablers that were identified through stakeholder feedback. This integrated approach facilitated nuanced understanding of stakeholder behaviour and enabled selection of well targeted strategies to address the diverse determinants influencing this implementation program [13, 47].

Another strength of this study is its exploration of mCIMT implementation in a generalist multidisciplinary ESD rehabilitation setting, an area that has not been previously researched [34, 48, 49]. While most CIMT research has been conducted in neurological rehabilitation environments [34, 48, 49], the availability of stroke-specific rehabilitation services beyond inpatient settings is limited, with most community-based stroke rehabilitation provided in generalist settings [50]. In these services, there is potential that therapists may de-prioritise complex interventions like mCIMT due to a lower proportion of stroke survivors on their caseload. This study provides valuable insights into how the intervention can be adapted for teams with generalist skills without compromising fidelity. This offers a model for broader dissemination beyond stroke-specific rehabilitation units and reinforces the importance of equitable access to evidence-based interventions, regardless of service configuration. The implications for rehabilitation services are substantial, particularly as the demand for multidisciplinary, community-based rehabilitation grows [50].

Conclusion

In conclusion, this study demonstrates the feasibility and acceptability of implementing mCIMT within a large, ESD rehabilitation service without access to additional resources. Despite initial challenges in adoption, the use of theory-informed behaviour change strategies and broad stakeholder engagement facilitated improvements in the rate of delivery of mCIMT. The findings highlight the need for continued efforts to address individual barriers (capability, skills, confidence, and motivation) and system-level constraints (workload pressures and organisational commitment). Given the high proportion of stroke survivors eligible, mCIMT should be advocated to become standard care in UL rehabilitation. Future research should examine the long-term sustainability of mCIMT delivery within the RITH service, and explore its impact across diverse rehabilitation contexts, including in regional and remote settings. Targeting policy-level change to prioritise the delivery of evidence-based stroke rehabilitation interventions like mCIMT requires health economics data from cost evaluation studies, which should be a key focus of future research. Achieving broad and sustained delivery of mCIMT has the potential to reduce disability and career burden, enhancing UL rehabilitation outcomes for up to one-third of all stroke survivors.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

AW- conceptualisation, methodology, investigation, resources, formal analysis, writing- original draft, writing- review and editing. EG- methodology, supervision, writing- review and editing. BS- methodology, supervision, writing- review and editing.

Funding

No funding was received for this study.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study fully adhered to the Declaration of Helsinki for research involving humans and human data. Ethics approval was obtained from the South Metropolitan Health Service (RGS0000003058) and Edith Cowan University Human Research Ethics Committees (2019 − 00989). Informed consent was obtained from all participants in the qualitative components of this study. For the quantitative data, which involved retrospective auditing of de-identified medical records, the South Metropolitan Health Service Human Research Ethics Committee granted a waiver of consent. All procedures were conducted with respect for privacy and data confidentiality.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.